Research On The Key Techniques Of SAR Interferometry For Regional Land Subsidence Monitoring

Land subsidence is gradual surface deformation procedure has been a major issueof geo-environment changing study. Regional subsidence occurred in the North ChinaPlan, the Yangtze Delta area and the Fenwei Graben indicates the contradiction ofground water supply and demand that is a command phenomenon in many plains,basins and delta area in China. Further development will increase the pressure onground water resources development and utilization which requires subsidencemonitoring should be a global perspective to research the geo-environmentbackground, dominant factors and consequence of subsidence, other than limited tosingle administrative divisions, and to find the optimized solutions for regionalgeo-environment problems between subsidence and ground water extraction. Regionalsubsidence over different administrative divisions provides urgent commands for highaccuracy and ability of large coverage and simultaneous measurement for themonitoring techniques. The conventional methodology such as precise leveling, highprecision GPS, bedrock bench mark and step bench mark are difficult to meet theabove requirements.SAR interferometry (InSAR) is a radar technique applied in geodesy and remotesensing, which exploiting phase differences of two or more SAR images returningto the satellite over the same area with slightly spatial baseline to generate maps ofsurface deformation or topography. The development of InSAR technique includingconventional Differential InSAR (D-InSAR) and Time Series InSAR such asPermanent Scatterer or Persistent Scatterer InSAR (PSInSAR) provides powerfulgeodetic tool for the monitoring of large scale surface deformation and regionalsubsidence. The technique can potentially measure changes in deformation over spansof days to years with centimeter or millimeter precision. It has been applied togeophysical monitoring of natural hazards, for example subsidence, earthquakes,volcanoes and landslides, and even in structural engineering, in particular monitoringof subsidence and structural stability. The PSInSAR consists in a series of numericalelaborations of radar satellite data, during which the data is studied statistically andnoise effects like disturbance due to the atmosphere, errors in the reference DEM, etc.are removed. At the end displacement values of points on the ground called PS areobtained. To apply the Technique it is necessary to have at least25-30radar imagesover the area of interest. However, in the case of less SAR images, how can wegenerate high accurate deformation map, and how to resolve key problems related tomonitoring of large coverage extends tens of thousands of square kilometers, and howto validate the precision and accuracy of InSAR measurement are problems can not beavoided for the operational application of InSAR techniques. This study aimed at the research of key techniques limited the operationalapplication of InSAR techniques for regional surface deformation monitoring. Theimprovement of the standard time series InSAR technique has been studied firstly. Tomeet the command of large coverage the integration method of multi-track and longstrip InSAR measurement has been established and the datum connection betweenground measurement and InSAR has been discussed. To the validation of InSARdeformation maps the intercomparion and cross validation have been carried out. Themajor innovative research is as follow.(1) Coherent Target InSAR (CTInSAR) techniqueBased on the standard PSInSAR and SBAS (Small Baseline Subset) technique animproved time series analysis method with the differential phase of the coherenttargets has been established in this work. The interferogram stacks are generated withmultiple master images other than the single one which applied in the standardPSInSAR. SAR images with perpendicular baseline smaller than300m and timeinterval with one or two years are combined for differential interferometric phasegeneration. To Coherent Target candidates (CTs) identification a multiple criterionaccording to the mean coherence, the sub-look correlation and the intensity stabilityhas been proposed. This method improves the accuracy of CTs selection and enablesthe ability to identify CTs in single SAR image as well as many ones. To estimateparameters such as linear velocity and height error of CT the two dimensionperiodogram spectral estimator has been applied with the unwrapped interferogramsstack. The Singular Value Decomposition method is adapted to estimate the nonlineardeformation and atmosphere of each image with the residual phase. After that thespatial and temporal filters are applied to get the nonlinear of each CT. Aboveprocessing chain degrades the necessary for large volume of images in the standardPSInSAR processing and can be carried out to data sets less than10-12images with2-3mm precision. T he cross validation with precise leveling verify that the meansquare error is3-5mm and meet the demand for regional subsidence monitoring.(2)Multi-track and long strip CTInSAR integration techniqueAimed to monitor subsidence occurs within very extended area a Multi-track andlong strip CTInSAR integration technique has been put forward in this study. Apatch-wise method has been introduced to process longer than500-700km long stripCTInSAR and a polynomial based fitting has been applied to remove the residualorbit fringes for large interferogram. To mosaic the multiple parallel long stripCTInSAR measurements the coordinate system transformation and the offsetcompensation have been introduced. The redundant observations in each adjacentparallel track have been inter-compared to validate the inconsistency of CTInSARmeasurements. The result indicates that the statistical square mean error is2.55mm and the difference of the same CT in adjacent track below5mm is greater than96%.(3) Datum connection between InSAR and ground measurementSince the InSAR derived deformation are referenced to a certain point in theinterferogram whose stability is unknown, the deformation velocity map estimated byCTInSAR are ‘floating’ and biased to ground leveling measurement which refer to aknown stable point. There exist an offset between the datum of InSAR and othermeasurements. To estimate the offset3kinds of ways have been introduced. The firstis the linear fitting of InSAR and leveling measurements with the nearly same point.The second estimate the difference between CTInSAR and leveling simultaneoussurveying over the same Corner Reflectors network. The third is based on the priorknowledge. These methods can be applied in the suitable case to accomplish thedatum connection.(4) Inter-comparison and cross validationBy completing the datum connection and coordinate transformation the CTInSARderived deformation map can be compared with the leveling measurements. In thiswork the inter-comparison are preformed for the CTInSAR measurements over theadjacent parallel tracks and the cross validation are carried out between CTInAR andleveling measurements. The validated results are listed in the previous sections andthe differences of the mean square error caused by different tie point searchingmethod are analyzed. To further compare the InSAR measurement and leveling thedistance related precision assessment method has been proposed and the result showthat the CTInSAR measurements reach the same precision with class leveling.5Applications and assessmentsThe previous key techniques have been applied for regional subsidencemonitoring in the NCP area. The CTInSAR results demonstrated the distinguishednonlinear subsidence and uplift related to seasonal changes of ground water extractionand recharge. The multi-track and long strip CTInSAR integration technique has beensuccessfully applied to get the subsidence in Beijing and Tianjin area. In addition, theabove technique has been performed to monitor subsidence along the JingJin HighSpeed Railway and subsidence induced risk migration assessment. The multipleapplications verified the accuracy and operational ability of this study.